Trigger Content and Function in Release 14
The purpose of this page is to outline the
goals for content and function of the offline trigger data at each level of the EDM (RDO/ESD/AOD/DPD/TAG) in Release 14. Size depends on menu and event type, so numbers given are estimates based on rel 13 default menu on ttbar events.
Trigger functions can be broken down into:
- full detailed analysis of trigger objects and all their properties
- re-running hypothesis algorithm to study optimizing algorithms and systematic uncertainty on efficiency
- did a trigger fire or not
RDO
...
ESD
Function: study all trigger objects in full detail, re-run hypo, retrieve trigger decision
Content: of ESD is expected to remain basically the same as in release 13.
Size: 200kB
AOD
Function: re-run hypo, retrieve trigger decision, other functionality?
Content: should AOD objects be 'thinned' to the minimum required to re-run hypothesis algorithms?
Size: 50kB
DPD
Function: retrieve trigger decision
Content: only what is necessary to retrieve the trigger decision (name or just chain ID?)
Size: 2kB
TAG
...
Cut Values EDM
Question: "What is the minimum possible trigger EDM size that one can retrieve the variables the hypo algos actually cut on?".
To answer the question, I'm asking each slice coordinator for a list which contains the minimum number of variables (float, int, etc.) required to run their hypo algorithms for L2 and EF.
L2 Summary Object |
Slice |
Respondent |
Content per Object |
# Objects per Event |
Size per Event (byte) |
Jet |
Brigitte Vachon |
1 float |
6 |
24 |
Electron |
Ricardo Goncalo |
10 floats |
9 |
360 |
Photon |
Valeria Perez |
6 floats |
16 |
384 |
Tau |
Olya Igonkina |
|
|
|
Muon |
Margherita Primavera |
L2-mufast: 8 float, 3 int L2-mucomb: 6 float L2-muIso: 5 float, 5 int |
2 |
216 |
MET |
Martine Bosman |
4 float |
1 |
16 |
Minbias |
William Bell |
|
Bphys |
Julie Kirk |
4 float |
2 |
32 |
Bjet |
Andrea Coccaro |
9 floats |
13 |
468 |
Cosmic |
Jamie Boyd |
|
|
|
Estimate for L2 Summary Object ~ 2 kB/event
"# Objects per Event" estimate comes from reading the number of collections per event from the checkFile.py output on top events with 13.0.30 default menu (obviously this is not exactly an "average" event)
"Size per Event" estimate is #objects times size of object (
bool = 1 byte, int = 4 byte, float = 4 byte, double = 8 byte
).
Working on an estimate for EF...
EF Summary |
Slice |
Respondent |
Answer |
Jet |
Brigitte Vachon |
1 float |
Electron |
|
|
Photon |
Valeria Perez |
|
Tau |
Olya Igonkina |
EFTau: 5 float, 2 int, 1 bool |
Muon |
Margherita Primavera |
EF-Moore: 3 float, 1 int EF-muid: 3 float, 2 int EF-comb: 8 float, 2 int, 3 long |
MET |
Martine Bosman |
|
Minbias |
William Bell |
|
Bphys |
Julie Kirk |
|
Bjet |
Andrea Coccaro |
|
Cosmic |
Jamie Boyd |
|
--
AndrewHamilton - 07 Jan 2008